Got to the bottom of a tricky fault :-)

[Circlotron]

Made some battery load testers at work to go inside these pieces of equipment we make and sell. Periodically applies a 2 amp load to a 24V SLA battery for 5 seconds and decides good or bad. Has a relay at the top of the load resistor and a mosfet at the bottom. Micro drives a CR network then to a 555 11 second timer that controls the relay. This is so that if the software goes haywire the relay will disconnect the load after 11 seconds regardless. Normally the mosfet is timed to allow the relay to switch under zero load so the contacts will last a long time.

At power up the micro checks that the mosfet switches on and off, the load resistor is not open cct, and the relay operates properly, all before any load is applied. If the 11 second 555 timer and consequently the relay doesn't release by 20 seconds it goes into fault mode. What was happening was the very first time the board was powered up it would fail this self test. (Edit - all four boards did exactly the same thing, but only once) What I finally figured was happening was the 555 is using a 100K and 100uF electro and this cap is about 4 or 5 years old, and it's leakage current was such that it cause the 555 to take more than the 20 second test threshold. Way more than it's intended 11 seconds. But only once! That was the tricky part. It was impossible to reproduce the fault. Only mulling over it while wearing my best thinking cap was I able to figure it out.

[moffy]

A nice piece of deduction, I guess the dielectric layer had deteriorated and needed a voltage to reform, hence the one off problem. You might have further problems down the line as the cap ages and loses value. Using an electrolytic cap in a critical timing application can be problematic.

[MarkT]

Using a CMOS 555, such as the 7555 allows much smaller cap values (i.e. larger R values), that might allow a non-electrolytic capacitor to be used...

[SeanB]

Yes replace the electrolytic with a tantalum capacitor, as the 555 is likely running off 5V best to use a 35V rated part there, which will not have any issues at all over the design life. Electrolytics leak, and in a 555 timer design fail with age, while the tantalum will survive for decades, provided you use modern ones from a reputable manufacturer.

[Siwastaja]

An excellent demonstration how "safety" features which add significant complexity tend to fail. This time, luckily, it failed in a safe way, and the good thing is, you did have the self-test feature. Somebody else would have been lazier and just trusted their independent 555 safety circuit without actually self-testing it.

Long timeconstant analog delay circuits are iffy. If you had batch of capacitors with higher leakage that would not self-heal on the first run, and if your self-test were not present or failed, then the safety circuit would be nonfunctional, too, and against MOSFET failure, it would have been better to simply control the relay from the microcontroller.

[coppercone2]

you just think you did. 2 months later you will realize that what you did is completely irrelevant and the problem is a burned wire or something

like it will go down without a fight 

[coppercone2]

Yes replace the electrolytic with a tantalum capacitor, as the 555 is likely running off 5V best to use a 35V rated part there, which will not have any issues at all over the design life. Electrolytics leak, and in a 555 timer design fail with age, while the tantalum will survive for decades, provided you use modern ones from a reputable manufacturer.

I made a prototype monitor type circuit for something or another a long time ago and I got told the same thing (luck had it that it WAS a tantalum)

might want to look at non conflict for engineering today. not sure what they would be but if someone is going to use this commercially its good to know

https://www.electronicspecifier.com/news/analysis/kemet-tantalum-capacitors-are-conflict-free